PMXBOT Log file Viewer

Help | Karma | Search:

#pypa logs for Friday the 17th of April, 2015

(Back to #pypa overview) (Back to channel listing) (Animate logs)
[11:26:32] <Wooble> I'm having reports that pip isn't finding suitable distributions of my library, which is at version "0.5a3". However, I can't reproduce this on my machine with pip 6. Has the need for the --pre flag been removed in later versions? I can't find this documented.
[11:27:28] <ronny> Wooble: whhich library is that?
[11:27:35] <Wooble> pycounter
[11:28:05] <Wooble> according to https://github.com/pitthsls/pycounter/issues/12 upgrading pip fixed it for the user.
[11:29:14] <ronny> Wooble: im pretty sure --pre is needed for your package unless it is explicitly requested at a exact version number
[11:29:30] <ronny> Wooble: any reason for not just ding a 0.5.0 release?
[11:29:30] <ionelmc> perhaps version normalization is what changed
[11:29:54] <Wooble> weird. I'm able to just pip install pycounter in a fresh virtualenv. I wonder if I just have it cached.
[11:30:33] <Wooble> ronny: probably not a good reason; I've been screwing with the API a bit, but I think I've removed all the cruft our local tool developed alongside it was using.
[11:31:50] <ronny> Wooble: before 1.x just increment the minor version on api changes
[11:32:42] <Wooble> fair enough.
[14:57:27] <seme> is there some standardised way to find out about and specify "optionals" when installing using pip? for example ipython has a number of optionals but there is no way to find out about them without looking to the source
[14:58:53] <ionelmc> seme: pypi has a json api, you could use that
[14:59:17] <seme> interesting... is anyone working on adding that kind of functionality to pip?
[14:59:32] <seme> I didn't see any open issues for that kind of thing
[14:59:52] <ionelmc> seme: dstufft might know
[15:00:13] <ionelmc> problem is that very few packages have the necesary metadata uploaded
[15:00:20] <ionelmc> and ipytohn is not one of them
[15:00:48] <seme> I'm guessing this is a bit of chicken or egg type problem... I would assume the functionality would need to be there before anyone would add that metadata to make use of it
[15:01:09] <ionelmc> kinda
[15:01:24] <ionelmc> pypi could at least automatically extract the metadata from the wheels
[15:01:54] <ionelmc> dstufft is working on warehouse (the next pypi)
[15:01:58] <ronny> pip has all metadata on install in any case
[15:02:22] <ionelmc> yeah but that means running setup.py
[15:02:23] <ronny> seme: am i understanding you correct, that you want potential extras to be listed on install?
[15:02:36] <ronny> ionelmc: so does pip install
[15:03:20] <ionelmc> yeah but then you don't expect that getting a list of extras for matplotlib would download and build numpy behind the scenes
[15:03:54] <ronny> seme: it might be helpfull to outline the use cases/users tories you have in mind, as yu can see, there are lots of funky edge-cases
[15:03:56] <seme> ronny: well I would imagine the workflow would be something like do a pip search for the package... then a pip info PACKAGE to see what optionals, detailed info and deps, etc. then finally you would just do the same thing you are doing now like pip install package[one, two, three]
[15:04:41] <seme> I mean I wouldn't want to change how pip works today but just give people the extra info they need to make choices like I want ipython[notebook] or ipython[all]
[15:04:55] <seme> right now I can't even tell what options are available without looking at the source
[15:05:29] <ronny> seme: i think its a good idea per se, but it needs some work figuring out the workflows, potential changes, things that can be done early, things that should wait for metadata 2.0
[15:05:48] <ronny> dstufft: ping?
[15:06:07] <seme> for example http://pastebin.com/C0WYT0sc
[15:07:34] <ronny> seme: there is a major difference tho, brew has all pacakge spec local for introspection ^^ pip has to inspect packages on pypi
[15:08:29] <seme> ronny: it could just be metadata that pypi provides to pip through the API
[15:09:27] <ronny> seme: then that would need to wait for metadata 2.0 in all packages, because currently its not programmatically availiable
[15:09:41] <seme> ronny: aah... ok
[15:10:00] <ronny> i.e. time to work propperly something between 3-5 years if we are lucky
[15:10:12] <seme> that kind of sucks
[15:10:21] <seme> ooh well :)
[15:12:40] <ronny> seme: its doable quicker, but that needs to have some kind of package downloading
[15:13:07] <ronny> seme: for anything that has wheels its luckyly easy, justdownloading the last few kb of a wheel usually contains the metadata
[15:13:36] <ronny> seme: and for sdist, one needs to download the package and take a look into the egg-info
[15:13:43] <ronny> seme: i think in general it canbe done reasonably
[15:13:57] <ronny> download caching and metadata introspection might need enhancements tho
[15:14:04] <ronny> (setup_requires is trickyfor example
[19:43:44] <frewsxcv> https://gist.githubusercontent.com/frewsxcv/8bcec4e3fad71c7cd3db/raw/ab3c2f05e4f8c9eca56e7052777edd26fb872d92/gistfile1.txt
[19:43:47] <frewsxcv> https://pypi.python.org/pypi/django-celery-transactions/0.3.0
[19:43:48] <frewsxcv> any ideas?
[19:45:05] <ionelmc> frewsxcv: you didn't upload anything
[19:45:29] <frewsxcv> i dont know what that means
[19:46:06] <ionelmc> frewsxcv: https://packaging.python.org/en/latest/distributing.html#upload-your-distributions
[19:46:17] <ionelmc> read the whole page actually
[19:46:32] <frewsxcv> i'm trying to download a library, not upload one
[19:47:04] <ionelmc> frewsxcv: aaah, seems the author forgot to make uploads
[19:47:07] <ionelmc> make a bug report
[19:47:25] <frewsxcv> why is someone allowed to upload nothing to pypi?
[19:49:01] <ionelmc> frewsxcv: cause no one punish him :-)
[19:49:16] <ionelmc> make a bug report, that's the punishment
[19:49:37] <frewsxcv> for pypi?
[19:49:54] <ionelmc> frewsxcv: https://github.com/fellowshipofone/django-celery-transactions/issues/new
[19:50:03] <ionelmc> describe your problem
[19:50:05] <frewsxcv> this seems like a usability flaw with pypi
[19:50:11] <ionelmc> press submit button
[19:51:03] <ionelmc> frewsxcv: there's the mailing list if you want to complain about that https://mail.python.org/pipermail/distutils-sig/
[19:53:14] <ronny> frewsxcv: its an artifct of history, the very first pypi has no uploads, only links
[19:53:38] <frewsxcv> is there any reason to keep that behavior around?
[19:53:55] <ronny> i think the next gen pypy no longer has it
[19:54:03] <frewsxcv> pypi?
[19:54:08] <ronny> frewsxcv: there is a replacement for the current coebase being made
[19:54:14] <ronny> frewsxcv: whops, yes, sorry, typo
[19:54:17] <ionelmc> frewsxcv: https://warehouse.python.org/
[19:54:40] <ionelmc> frewsxcv: you could check https://github.com/pypa/warehouse/issues/
[19:55:11] <ionelmc> maybe even make a report so people are aware of your perspective
[19:55:43] <frewsxcv> is it just a new frontend?
[19:56:13] <ionelmc> it's gonna fix some of the long standing issues pypi has
[19:56:36] <ionelmc> i'm not very familiar with the details
[19:56:58] <ionelmc> frewsxcv: dstufft is working on it, maybe you could talk to him
[19:59:59] <frewsxcv> thanks for letting me know about my specific issue
[20:00:49] <ionelmc> frewsxcv: it's just one of the very many mistakes that a package author can make
[20:01:38] <frewsxcv> yes, even as a seasoned python developer, i still feel like i dont grasp all the packing options/configurations
[20:01:48] <frewsxcv> not as easy as: npm publish or cargo publish
[20:02:47] <ionelmc> you're not alone